review-board

A Claude Code plugin providing 25 specialized review agents backed by 42 modular skills, covering problem analysis, code quality, testing, security, compliance, observability, and architecture review.

Each agent has a narrow scope and hard boundaries — they're meant to be composed, not asked to do everything at once.

Installation

This plugin is distributed through its own Claude Code marketplace. Inside Claude Code:

/plugin marketplace add liambeeton/review-board
/plugin install review-board@review-board

Pull future updates with:

/plugin marketplace update review-board

Alternative: install from a local clone

If you'd rather work from a local checkout (e.g. to hack on the agents):

git clone https://github.com/liambeeton/review-board.git ~/code/review-board

Then inside Claude Code:

/plugin marketplace add ~/code/review-board
/plugin install review-board@review-board

Once installed, the agents and skills are available in any Claude Code session.

Usage

Agents are invoked through the Agent tool. You can either ask Claude to use a specific agent by name, or describe the kind of review you want and let Claude pick the right one — every agent has a description field that Claude uses to route work.

> Run the code-smell-detector on src/payments/
> Have the security-assessor look at the new auth flow
> Use the system-explorer to walk me through this codebase

Agents produce structured reports (Markdown or JSON, depending on the agent). Sample outputs for most agents live in docs/<agent-name>/examples/.

Agents

Problem framing & exploration

• problem-analyzer — decomposes a problem into subproblems and surfaces contradictions before any implementation
• system-explorer — walks through a codebase and produces a guided tour
• system-auditor — runs a compliance-driven audit against one or more frameworks (SOC 2, ISO 27001, PCI DSS, DORA, etc.)

Code quality & refactoring

• clean-coder — applies clean-code guidance to a change
• code-smell-detector — identifies smells, anti-patterns, and SOLID/GRASP violations
• refactoring-advisor — proposes Fowler-style refactoring sequences
• dead-code-detector — finds unreachable or unused code
• consistency-checker — surfaces conventions that drift across the codebase
• cognitive-load-analyzer — measures cognitive load index per module
• legacy-code-analyzer — assesses legacy modules for risk and seams
• ownership-analyzer — surfaces code ownership and bus-factor risk

Testing

• test-design-reviewer — evaluates test quality and detects test theatre / mock anti-patterns
• atdd-developer — drives outside-in acceptance test development
• user-story-writer — turns problem analyses into INVEST-style stories with acceptance criteria

Architecture & domain

• ddd-assessor — evaluates domain-driven-design alignment
• api-design-reviewer — reviews API surface and contract design
• data-layer-reviewer — reviews persistence and data access concerns

Security & compliance

• security-assessor — application-level threat and vulnerability review
• dependency-auditor — third-party package and supply-chain hygiene
• accessibility-assessor — WCAG-driven accessibility review

Operational readiness

• devops-evaluator — delivery practices and CI/CD posture
• observability-assessor — logging, metrics, and tracing coverage
• error-handling-reviewer — failure-mode hygiene
• concurrency-analyzer — race conditions, contention, and ordering issues

Documentation

• documentation-reviewer — README, ADR, and inline documentation quality

Skills

Agents are backed by 42 reusable skills under skills/ — Fowler refactoring catalogs, smell taxonomies, comprehension models, audit methodology, evidence and reporting schemas, CLI scoring formulas, compliance framework knowledge bases (CCPA, CMMC, DORA, FedRAMP, GDPR, HIPAA, ISO 27001, NIS2, NIST CSF, PCI DSS, SOC 2, SOX), and WCAG criteria.

Skills auto-load when an agent (or any Claude Code session) needs the knowledge — you don't invoke them directly.

Repository layout

.claude-plugin/        plugin and marketplace manifests
agents/                25 agent definitions (Markdown with YAML frontmatter)
skills/                42 skill bundles (SKILL.md, optional lib/)
docs/<agent>/          per-agent README, research notes, sample outputs

License

See repository for licensing details.